First, the uncomfortable finding about ChatGPT
Of the five engines, ChatGPT cites the fewest sources by far: 96 citations across our 30 questions, versus 386 for Claude, 311 for Gemini and 291 for Perplexity. When it does cite, it prefers editorial review media — TechRadar alone took 11 of its 96 citations, more than any vendor's own site.
Now the part most guides miss. The tool ChatGPT recommended most in our category was Ahrefs — named in 40% of its answers — while earning only 4 of those 96 citations. Ahrefs is the biggest, oldest brand on the list. Meanwhile the specialist tools that dominate the other engines' answers scored lower on ChatGPT (Otterly 27%, Peec 23%, Profound 17%).
Each engine eats differently — the per-engine playbook
Here is what each engine's answers actually cited in our scan, and the action that follows from it:
| Engine | Citations | What it cited most | So you should… |
|---|---|---|---|
| ChatGPT | 96 | TechRadar 11×; little else concentrated | Pitch editorial reviews; build brand familiarity; be patient |
| Perplexity | 291 | HubSpot blog 22×, Semrush 13× | Get included in the big marketing listicles it already cites |
| Gemini | 311 | YouTube 20×, Reddit 18× | Real video content + genuine community participation |
| Google AI Overviews | 62 | YouTube 10× | Video + classic SEO fundamentals — it answers least often |
| Claude | 386 | Wikipedia 10×, vendors' own sites (e.g. otterly.ai 10×) | Reference-grade, factual pages on your own domain |
Counts are from the Nov 16, 2026 scan: every citation URL in every answer to the 30 buyer questions, host-level. A channel that moves one engine can do nothing for another — this table is why "one weird trick" articles disappoint.
The playbook, in order
- Measure your baseline first. You cannot steer what you don't measure, and your intuition will be wrong about which engines already know you. (Our own baseline is public: 0% of 138 answers, 0 of 1,146 citations. We publish it because the detector is cleanest at zero.)
- Publish the pages AI wants to cite — on your own domain. In our category, 11 of the 20 most-cited domains are vendors' own sites. The engines demonstrably cite vendor content when it's specific and useful: question-phrased FAQs, measured comparisons, honest data. Generic "ultimate guides" are not what the citations show.
- Knock on the doors already cited. A specific set of listicles and review pages feeds your category's answers — find which ones the engines cite, and pitch inclusion there. This is faster than waiting to be discovered: you're adding your name to pages the engines already trust.
- Work the per-engine channels from the table. Video and Reddit for Gemini, major listicles for Perplexity, editorial reviews and brand-building for ChatGPT, reference content for Claude.
- Re-measure weekly and expect weeks, not days. Entry is genuinely possible: in our own category's 60-day window, Peec AI went from 0% to 43% of answers. That's the proof the game is open — and the reason to track it with the same questions every week rather than one-off checks.
What measurably doesn't work
Paying for ChatGPT Business Hub verification (as a visibility tactic)
When we asked the engines directly whether verification improves visibility, ChatGPT said yes — citing zero sources. Perplexity, Gemini and Claude all said no, each citing OpenAI's own documentation, which states verification does not change ranking. Every engine that retrieved sources said no. We're testing this in public with a verified profile and weekly scans — the full write-up is in our Business Hub report.
Chasing citations without a brand
Being cited gets your content into answers, not necessarily your name into recommendations. ChatGPT's own numbers above make the point: the most-recommended tool took 4 citations; a review site with nothing to sell took 11. Citations are a channel — the recommendation follows the brand.
Where we stand, honestly
We sell an AI-visibility tracker and our measured share is currently 0%. This guide is literally our own playbook, published, with the baseline attached — and we re-scan the same 30 questions weekly, so whatever happens next is on the record. If the playbook works, you'll see it in the data; if not, you'll see that too.
Method & limitations
- 30 buyer-intent questions about AI visibility / AEO tools, asked on ChatGPT, Perplexity, Gemini, Claude and Google AI Overviews, US market, Nov 16, 2026. Multi-run consensus per cell; 138 of 150 cells returned an answer (Google AI Overviews answered 18 of 30).
- Citation counts are host-level tallies of every citation URL in every answer (1,146 total). Recommendation share = share of answered cells naming the brand, exact-token matching.
- One category, one market, one month. Source preferences shift as engines update; we re-scan weekly and will correct this page if the picture changes.
- We compete in this category. Every number here comes from the same raw data as our published comparisons, including our own 0%.
Step 1 is free: see your baseline.
One real buyer question across five engines, free, no email — including which sources feed the answers when AI doesn't name you. Then $500 for the full one-time diagnostic, or $200/mo for weekly tracking.